Figure 4

Unsupervised clustering based on gene expression profiling of benign prostatic epithelial, primary prostate cancer, and metastatic prostate cancer cells laser capture microdissected from 26 cases. Results of unsupervised clustering using the dChip PM-only model (A and B) or the PM/MM difference model (C and D) to calculate expression values, which were (B and D) or were not (A and C) log transformed. Approximately 1500–2000 genes with the highest standard deviation/mean (not log transformed) or the highest standard deviation (log transformed) were used in the analysis. Green: benign; teal: benign prostatic hyperplasia; blue: Gleason pattern 3; orange: Gleason pattern 5; pink: metastatic prostate cancer; normalizer: normalizer sample (Gleason pattern 4, see Results).